A broad, locally hosted MyFitnessPal MCP server with 20 tools covering diary reads and writes, food search, custom-food management, body measurements, exercises, nutrition goals, water and fasting writes plus nutrition reports, with Chromium browser auto-discovery for authentication. It is not an official project: MyFitnessPal has no public API, reads are scraped, and writes rely on undocumented internal endpoints that can change at any time; the README also states plainly that legacy form login mostly fails after MFP's move to NextAuth, so browser-session cookies are the reliable path. Suitable for individuals who accept those constraints; not suitable where official support or policy certainty is required, or where storing session cookies locally is unacceptable.
A locally hosted Model Context Protocol server that lets AI assistants such as Claude interact with your MyFitnessPal data: food diary, exercises, body measurements, nutrition goals, water intake and intermittent fasting. MyFitnessPal has no public API: reads are scraped from the website via python-myfitnesspal, diary writes go through MFP's internal v2 JSON API (the same one its web client uses, authenticated with your existing session token), and custom-food writes use a different family of cookie-authenticated web endpoints. The server runs locally over stdio; data is only transmitted between your machine and myfitnesspal.com.
{
"mcpServers": {
"myfitnesspal": {
"command": "/Users/yourname/myfitnesspal-mcp-python/venv/bin/python",
"args": ["-m", "mfp_mcp.server"]
}
}
}